Continuous online process analytics with multiplexing gas chromatography by using calibrated convolution matrices

Abstract

The development of fast and precise measurement techniques for process analytical technology is important to operate chemical processes safely and efficiently. For quantitative measurements of multiple components at a trace level, often gas chromatographic methods are used which have a response time of several minutes or of up to one hour. For fast changing processes, this can be too slow for efficient control. For reducing the dead time of a control loop by increasing the measurement frequency, a multiplexing gas chromatography (mpGC) technique for a chromatographic system exhibiting a systematic non-linear response has been developed. For mpGC, superimposed chromatograms are measured by injecting consecutive samples before all components of previous samples have eluted from the column. The deconvolution of a superimposed chromatogram yields a computed chromatogram which is an average over the single chromatograms forming the superimposed chromatogram. Such a computed chromatogram typically shows so called correlation noise depending on the degree by which the single chromatograms forming the superimposed chromatogram will differ from each other (non-linear response). A technique is presented to calibrate the convolution matrix in order to suppress correlation noise introduced by systematic errors of the chromatographic system. The remaining correlation noise in the computed chromatogram is then exclusively caused by changing concentrations in the sample stream. For the method presented here, the sample is injected five times during the run time of a single chromatogram. The computed chromatogram is obtained three times within this timespan while representing each time an averaged chromatogram over the last five injections. Therefore, the sample throughput is increased by a factor of three compared to conventional GC.